Create Page Inputs & Settings

The Create page has the creator on the left and song results on the right. Start at the top, choose a model and mode, complete the visible fields, check the credit cost, then choose Create.

The full desktop Create page annotated with five areas: mode, model, inputs, credits with Create, and the right-hand song gallery. — Read the workspace in order: 1 mode, 2 model, 3 visible inputs, 4 cost and Create, then 5 processing and completed songs.

1. Choose a music model

Open the model button at the top of the creator. Each card shows the current version, a short description, availability, and generation cost.

Model family	Best starting point	Main behavior
AIMusicGen V2–V7	General song creation and the widest set of song actions	Easy or Custom creation; current versions also support compatible Audio and Reference paths
MiniMax Music 3.0	A song with your own lyrics and a chosen length	Custom lyrics are required; duration is 1–300 seconds; Instrumental and Reference are unavailable
ElevenLabs Music V2	One full prompt or a song planned section by section	Duration is 3–600 seconds; Sections can control text, length, styles, and context for each part

The model selector comparing the three current song model families. — Read the active model card before creating. The model changes available fields, limits, output count, and price.

AIMusicGen song generation costs 4 credits and returns two versions. MiniMax uses 1 credit for each requested second. ElevenLabs uses 10 credits for each started minute, with a 10-credit minimum. Always use the live Credits required value as the final price because model availability and pricing can change.

Changing the model can change field limits and available controls. Recheck the visible form after every switch. For help choosing, see Choose a Music Model.

2. Choose Easy or Custom

Easy

Use Easy when one description is enough.

Song Description accepts up to 500 characters.
The dice loads a complete example.
A suggestion replaces the current description, so copy any text you want to keep first.
Instrumental asks for music without vocals.
Lyrics carries the current idea into Custom so you can add or edit exact words.

Custom

Use Custom when you need exact Lyrics, Style, Title, visibility, vocal preference, saved snippets, Audio, Reference, or model-specific duration and section settings.

3. Complete Lyrics and Style

In Lyrics, type your own words or use Generate Lyrics. When text already exists, use Optimize to improve it. Basic and Advanced versions can be edited before Use this lyrics applies them. Save reusable text in Saved Lyrics.

In Style, describe genre, mood, tempo, instruments, voice, arrangement, and production. Use Quick styles or + More for tags, Enhance to expand a short idea, and Advanced options → Exclude styles for sounds you do not want. Save a useful prompt in Saved Style.

Both editors show the active character limit and can be resized. More guidance:

Example: put one idea into the correct fields

Suppose you want a late-night R&B song with exact chorus words.

In Easy, the whole request can stay in Song Description:

Late-night alternative R&B about calling someone after a long separation, 76 BPM, soft electronic drums, warm sub bass and muted guitar, intimate male vocal, restrained verse opening into a wider hopeful chorus, spacious mix.

In Custom, separate the exact words from the sound:

Lyrics

[Verse - close and restrained]
The room is quiet, half past two
I trace the words I never sent to you

[Chorus - wider harmonies]
If I call, will you stay on the line?
Meet me where the city loses time

Style

Late-night alternative R&B, 76 BPM laid-back pocket, soft electronic drums, warm sub bass, muted guitar, intimate male lead, restrained verse, wider chorus with close harmonies, spacious delay and low warm reverb.

Exclude styles

shouted vocals, rock guitars, bright dance drop, orchestral strings

Title

Stay on the Line

Leave Instrumental off because the request includes vocals. This separation gives each field one job: Lyrics says what is sung, Style says how it sounds, and Exclude styles names only unwanted elements.

Model-specific input examples

MiniMax Music 3.0: enter the complete lyrics in Lyrics, add a Style such as warm acoustic soul, brushed drums, intimate female vocal, then choose a duration such as 90 seconds.
ElevenLabs Prompt: write one complete request such as Three-minute cinematic synth-pop song, restrained opening, driving second verse, large final chorus, female lead, clean ending, then choose the total duration.
ElevenLabs Sections: create separate parts such as a 12-second sparse Intro, a 30-second Verse with close vocals, and a 24-second Chorus with full drums and stacked harmonies. Use each section’s Text, Styles, Avoid, and Context fields only for that part.

4. Add a source when needed

Audio lets you upload a file, save it without generating, or start a compatible Cover, Extend, Add Instruments, or Add Vocals action. See Upload Audio & Choose an Action.
Reference lets you reuse a saved song or voice identity. It is shown only when the active model and action support it. See Song & Voice References.

When you open a source action, the form changes to the fields required by that action. If a field disappears, it is no longer part of the request.

5. Check the remaining settings

Vocal Gender: Female or Male is an overall preference. Click the selected option again to clear it.
Visibility: Public can be shared. Private requires an eligible plan and can also be changed after generation.
Title: names the result; the limit follows the active model.
Duration: appears for MiniMax and ElevenLabs.
ElevenLabs Prompt / Sections: Prompt uses one full instruction. Sections lets you add, reorder, duplicate, or delete up to 30 parts. Each section is 3–120 seconds, and the full song must remain within 3–600 seconds. A section can include text, Styles, Avoid, and Context.

6. Check credits and create

The footer shows Credits required. Create stays disabled until every visible required field is valid.

Before clicking Create, confirm:

the active model and mode;
Lyrics or Instrumental;
Style and Exclude styles;
any Audio or Reference source;
duration, visibility, vocal preference, and title;
the displayed credit cost.

After submission, new rows appear in the results panel. AIMusicGen returns two rows for a normal song request; MiniMax and ElevenLabs return one. Processing rows update automatically, so do not submit the same request again just because the result is not immediate.

7. Review the result

Play the whole song, including its ending. A completed row can be renamed, favorited, downloaded, shared, moved between collections, or given a private note. Open More actions for compatible editing tools such as Extend, Cover, Replace Section, Add Instruments, Add Vocals, or input reuse.

If the result is close, change one part of the prompt and create a new version. If only one area is wrong, use a local editing action instead of rebuilding the whole song. See Best Practices for a repeatable review loop.